Consider transposed input matrix

[kieranrcampbell]

I originally wrote that if an SCE isn't provided and a user provides a count matrix then effectively it should be cell by gene, ie

t( counts( sce ) )

as input, out of a pedantic need that dim 1 should be samples (cells). However, this is probably over confusing and we should change it to just gene by cell (ie counts(sce) equiv)